Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Selective descriptor pruning for QSAR/QSPR studies using artificial neural networks.

Joseph V Turner1, David J Cutler, Ian Spence

  • 1Faculty of Pharmacy, The University of Sydney, NSW 2006, Australia. jvturner@pharm.usyd.edu.au

Journal of Computational Chemistry
|April 15, 2003
PubMed
Summary

Artificial Neural Networks (ANNs) effectively select optimal descriptors for quantitative structure-activity-property relationship (QSAR/QSPR) studies, even with noisy data. This method identifies meaningful variables among numerous false ones, improving QSAR/QSPR model development.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Modeling rare coding variation on chromosome X provides insight into the genetics and differential sex prevalence of autism spectrum disorder.

medRxiv : the preprint server for health sciences·2026
Same author

Heritability of Long-Term Complications in Classic Galactosemia.

Journal of inherited metabolic disease·2026
Same author

DASS-21 Stress, but Not Anxiety or Depression, Is Associated with Premenstrual Stress.

Journal of clinical medicine·2025
Same author

A powerful framework for differential co-expression analysis of general risk factors.

Bioinformatics (Oxford, England)·2025
Same author

Distinguishing syndromic and nonsyndromic cleft palate through analysis of protein-altering de novo variants in 818 trios.

American journal of human genetics·2025
Same author

Multiancestry brain pQTL fine-mapping and integration with genome-wide association studies of 21 neurologic and psychiatric conditions.

Nature genetics·2025

Area of Science:

  • Computational Chemistry
  • Cheminformatics
  • Machine Learning in Drug Discovery

Background:

  • Descriptor selection is a critical challenge in quantitative structure-activity-property relationship (QSAR/QSPR) modeling.
  • Artificial Neural Networks (ANNs) are powerful tools for QSAR/QSPR but underutilized for descriptor selection.
  • Handling large datasets with numerous input variables, including noisy and random descriptors, complicates model development.

Purpose of the Study:

  • To investigate the efficacy of Artificial Neural Networks (ANNs) for selecting optimal descriptors in QSAR/QSPR studies.
  • To evaluate ANN performance with diverse data types (clean, noisy, random) and data patterns (linear, nonlinear, periodic).
  • To determine the influence of the rho parameter on network architecture and its relation to sample size.

Related Experiment Videos

Main Methods:

  • Utilized ANNs for descriptor selection from a large pool of input variables.
  • Examined the impact of clean, noisy, and random descriptors on synthetic and real QSAR/QSPR datasets.
  • Employed a signal-to-noise ratio method for identifying the optimal descriptor set.
  • Analyzed the relationship between sample size, network architecture (rho parameter), and data characteristics.

Main Results:

  • ANNs successfully identified an optimal set of descriptors, demonstrating robustness against noisy and random inputs.
  • The signal-to-noise ratio method proved effective for determining the best descriptor subset.
  • Optimal rho parameter values varied depending on the specific data type and QSAR/QSPR problem.
  • ANNs effectively distinguished relevant descriptors from a large number of irrelevant ones.

Conclusions:

  • ANNs offer a viable and powerful approach for optimal descriptor selection in QSAR/QSPR studies.
  • The signal-to-noise ratio method, combined with ANNs, enhances the reliability of descriptor selection.
  • Understanding the role of the rho parameter is crucial for optimizing ANN performance in descriptor selection across different datasets.